Claude Code
Claude Code security review: the command, the GitHub Action and what they leave to you
A plain guide for engineering leads and security teams. Every fact about the tools comes from Anthropic's documentation and Anthropic's published repository, read on 11 November 2026. Settings change, so check the cited pages before you act.
Claude Code security review is a way to have Claude check code changes for security problems, and Anthropic offers it two ways. The /security-review command, built into Claude Code, analyzes the changes on your current branch for vulnerabilities such as injection, authentication issues and data exposure. A separate GitHub Action published by Anthropic reviews pull requests and comments on the lines it flags. Anthropic's repository says the Action is not hardened against prompt injection and should only review trusted pull requests, so a human reviewer still decides what is merged.
plain answer
What is Claude Code security review?
Claude Code security review means asking Claude to read a code change and report security vulnerabilities in it. Anthropic's Claude Code command reference describes the /security-review command as analyzing the changes on your current branch for security vulnerabilities, by reviewing the difference between your branch and the default branch of the origin remote. It names injection, authentication issues and data exposure as examples of what it looks for. The same page says the command needs an origin remote.
Anthropic also publishes a GitHub Action, claude-code-security-review, described in its repository as an AI-powered security review that analyzes code changes for security vulnerabilities and comments on pull requests. The repository says the command and the Action offer the same security analysis, one inside your Claude Code session and one in your pull request workflow.
Sources: Anthropic, Claude Code Commands (https://code.claude.com/docs/en/commands) and Claude Code Security (https://code.claude.com/docs/en/security); Anthropic, claude-code-security-review repository README (https://github.com/anthropics/claude-code-security-review), all read 11 November 2026.
how to
How do you run a Claude security review on your code?
To run a Claude security review on a branch, open Claude Code in the repository and run /security-review. Anthropic's repository says this performs a security review of all pending changes. The command reviews your branch against the origin default branch, so a repository with no origin remote will not work, and Anthropic's error reference has a section for a failure with an ambiguous argument message.
To review pull requests automatically, the repository shows a workflow file that runs on pull requests, checks out the pull request code and calls the Action with a Claude API key stored as a secret. The Action needs permission to write pull request comments and to read repository contents. The README says the API key must be enabled for both the Claude API and Claude Code usage.
Anthropic's repository says a project can copy the repository's security-review.md file into its own .claude/commands folder and edit it, for example to add its own directions. Neither page states which Claude plans include the /security-review command, so check your plan with Anthropic.
Sources: Anthropic, Claude Code Commands (https://code.claude.com/docs/en/commands); Anthropic, claude-code-security-review README (https://github.com/anthropics/claude-code-security-review), read 11 November 2026.
how to
What does the GitHub Action check, and how does it report?
The Action's README lists these features: it scans only the changed files in a pull request, comments on the pull request with its findings, works with any programming language, and filters findings to reduce noise. Its workflow has five steps: it reads the change, examines the code in context, writes findings with explanations, severity ratings and remediation guidance, filters out low impact findings, and posts comments on the specific lines.
The README lists the kinds of problems it looks for, including injection attacks, authentication and authorization, data exposure, cryptographic issues, input validation, business logic flaws, configuration security and supply chain risk. It also says what the false positive filter leaves out by default: denial of service, rate limiting concerns, memory or CPU exhaustion, generic input validation without proven impact, and open redirects. That list matters, because a quiet report does not mean those problems are absent.
Teams can tune this. The README describes inputs for excluding directories, for custom scan instructions and for custom false positive filtering instructions, each given as a text file.
Sources: Anthropic, claude-code-security-review README (https://github.com/anthropics/claude-code-security-review), read 11 November 2026.
security
What should a team decide before relying on a Claude security review?
Start with trust. Anthropic's README says the Action is not hardened against prompt injection attacks and should only be used to review trusted pull requests. It recommends the repository setting that requires approval for all external contributors, so the workflow runs only after a maintainer has reviewed the pull request. A team that accepts outside contributions should decide who approves those runs.
Next, ownership. Anthropic's Claude Code security page says you are responsible for reviewing proposed code and commands for safety before approval, and that no system is completely immune to all attacks. The review is a second reader, not a gate. Decide who reads the findings, who can dismiss one and where that decision is recorded.
Then, keys and data. The Action needs a Claude API key stored as a repository secret, so decide who owns that key and who can change the workflow. The Action asks Claude to analyze your code changes, so check Anthropic's commercial terms and privacy pages for how it is handled before you point the Action at a sensitive repository.
Sources: Anthropic, claude-code-security-review README (https://github.com/anthropics/claude-code-security-review); Anthropic, Claude Code Security (https://code.claude.com/docs/en/security), read 11 November 2026.
plain answer
How does this differ from Claude Code review?
Claude Code security review is about security problems in a change. Claude Code review, Anthropic's separate pull request review service covered on our Claude Code review page, reads each change for correctness and says a review with no findings does not mean the code is error-free or secure. Anthropic's command reference also lists a /code-review command for correctness bugs next to /security-review for vulnerabilities.
In practice a team may use both, with different owners. Our Claude security page covers the wider question of where Claude runs and what it may touch.
Sources: Anthropic, Claude Code Commands (https://code.claude.com/docs/en/commands), read 11 November 2026; Kriv AI, Claude Code review page, read from the Anthropic Code Review documentation on 11 November 2026.
kriv role
Where Kriv helps
Kriv AI trains engineering teams to use Claude Code in their own codebase, including review discipline, CI integration and a governance module. For security review that means agreeing who owns the findings, which pull requests may run the Action, and what a human reviewer still checks. Kriv AI has achieved the Select partner tier in the Claude Partner Network Services Track.
Kriv did not build these tools and does not replace Anthropic's documentation. Kriv does not perform penetration testing or certify that code is secure. Your own engineers and security team hold the access and the final say.
limits
What this page could not confirm
We did not run the command or the Action. The pages we read show no publication dates, and the repository shows no tagged releases. We did not confirm which Claude plans include /security-review, what the Action costs, or how Anthropic handles code sent for analysis. For anything not cited above, check Anthropic's current pages.
Related resources
Continue exploring
Straight answers
Frequently asked questions about Claude Code security review: the command, the GitHub Action and what they leave to you
What does the /security-review command do?
Anthropic's command reference says it analyzes the changes on your current branch for security vulnerabilities, comparing your branch with the origin default branch, and names injection, authentication issues and data exposure as examples (read 11 November 2026).
Is there a Claude Code security review GitHub Action?
Yes. Anthropic publishes claude-code-security-review, which reviews pull requests for security vulnerabilities and comments on the lines it flags. Its README says it should only be used to review trusted pull requests (read 11 November 2026).
Does a clean security review mean the code is secure?
No. Anthropic's Claude Code security page says no system is completely immune to all attacks and that you remain responsible for reviewing code before approval, and the README lists problem types the filter leaves out by default (read 11 November 2026).
Can I change what the Action looks for?
Yes. The README describes inputs for custom scan instructions, custom false positive filtering instructions and excluded directories, each supplied as a text file or a directory list (read 11 November 2026).
Can Kriv set up security review for our team?
Kriv AI trains engineering teams on Claude Code, including review discipline and a governance module, and helps agree who owns findings. Your own security team keeps the final say. Kriv does not certify code as secure.
Talk to the team that would do the work
Bring your requirements to a working session with the person who'll actually deliver.
Book a Discovery Call